Demographic parameters are crucial aspects of animal ecology and life histories, and one of the main tools for biodiversity conservation (Vitt and Caldwell 2009). Robust data on turtle demography are critical for effective management and conservation, especially if impacts on population survival, such as anthropogenic factors, are considered (Gibbons et al. 2001). For management success, the complexity of turtle population characteristics demands focus on selected demographic data and also great care in analyzing estimated parameters. Indeed, population models now available demand greater responsibility among researchers to perform the necessary assessments so that the most appropriate models are selected (Amstrup et al. 2005; Cooch and White 2008). At the same time, to predict population abundance or viability with high precision is a difficult task, especially given the insufficient information on species life histories (Huston 2002).

About 17% of the 328 known turtle species occur in South America, including 8 families. The Chelidae has the largest geographical distribution and is the most species-rich family (19 of 22 described South American species occur in Brazil) (Souza 2005). Knowledge of the status of non-Amazonian turtles in Brazil is sparse, especially for the Cerrado biome, which is one of the 34 world biodiversity hotspots (Myers et al. 2005). Six species of chelids occur in the Cerrado (Colli et al. 2002): Acanthochelys spixii, Chelus fimbriatus, Mesoclemmys gibba, Mesoclemmys vanderhaegei, Phrynops geoffroanus, and Platemys platycephala. When considering the incipient knowledge about non-Amazonian chelids (Souza and Molina 2007), life history studies conducted under natural conditions are important to inform conservation and management programs (Souza 2004). A demographic analysis of Mesoclemmys hogei, the most threatened Brazilian freshwater turtle (Tortoise and Freshwater Turtle Specialist Group 1996b), highlighted the importance of habitat preservation and concluded that low survival rates are related to predation at initial development stages (Moreira 2002). A long-term study on Hydromedusa maximiliani estimated an annual population growth of 1.012, which indicates a healthy population because of the level of protection of the study site (Souza 1995; Souza and Abe 1997; Souza and Martins 2006; Martins and Souza 2009).

Although there are no demographic studies on Cerrado turtles, amphibian and lizard populations show a tight association between demography and climate, a consequence of cyclic reproduction (e.g., Wiederhecker et al. 2003; Vasconcellos and Colli 2009). Demographic studies of Cerrado species are warranted because 1) at least 80% of the biome has been modified by human activities (Dias 1994); 2) until 2008, the Cerrado had lost about 48% of its native landscape, and only 2.89% of its territory is adequately protected (Ministério do Meio Ambiente 2009); and 3) degradation of aquatic ecosystems in the Cerrado is ongoing (de Sá et al. 2003; Nogueira et al. 2010).

Acanthochelys spixii (black spiny-necked swamp turtle) (Fig. 1A, B), popularly known in Brazil as “cágado-preto,” is a small (maximum carapace length around 170 mm) semi-aquatic turtle that inhabits freshwater and terrestrial habitats associated with temporary swamps, slow moving waters, heavily vegetated aquatic bodies, and small rivers (D'amato and Morato 1991; Tortoise and Freshwater Turtle Specialist Group 1996a). In captivity, these animals display little activity during the cold season, which suggests hibernation in the wild (Molina and Rocha 1990). Acanthochelys spixii has a broad distribution in South America, and occurs from Uruguay and Argentina through the Brazilian states of Rio Grande do Sul, Paraná, São Paulo, Minas Gerais, and Distrito Federal (its northernmost limit) (Rhodin et al. 1984; Brandão et al. 2002). In Brazil, A. spixii inhabits areas with severe winters in the southern states to areas with a mean annual temperature of approximately 22°C. Suitable habitats for the species include contact zones between open (e.g., Cerrado) and closed (Atlantic rainforest) formations (Souza 2005). The species was considered vulnerable in Uruguay, because of its near-threatened International Union for the Conservation of Nature (IUCN) status, low abundance, and vulnerability to environmental changes (Carreira et al. 2007). In 1998, the Brazilian state of São Paulo listed A. spixii as vulnerable (State Decree 42.838/98). Herein, we estimate demographic parameters and model the effects of sex and climatic variation in a population of A. spixii from the Cerrado biome of central Brazil, based on 4 years of mark–recapture research. We hypothesize a significant association of demographic and climatic parameters, given previous studies with other herpetological fauna in the region.

View Figure

• Download as Powerpoint
• Download Figure

METHODS

RESULTS

During the 48 months of the study, 56 adult turtles (23 females and 33 males) were individually marked and captured 174 times, with 43 individuals being recaptured at least once. The sex ratio among marked animals was slightly male biased, at 1.43∶1, which did not differ significantly from 1∶1 (χ²_[1]  =  1.79; p  =  0.18). Adult females were significantly heavier and larger than males (Table 1). Among 174 capture events, leeches (Batracobdella sp.) were observed on turtles on 102 occasions; 1 male turtle was captured with 38 leeches attached to its body. The water level varied up to 116 cm and was positively associated with total rainfall in the previous month (r_S  =  0.66, p < 0.001).

Table 1 Body measurements of adult Acanthochelys spixii in Parque Nacional de Brasília, Brazil, according to sex, with analysis of variance results. Data are mean ± SD.

View Fullscreen

The overall goodness-of-fit test detected no general significant deviation from fit of the global model (Φ_s*t, p_s*t) to the CJS model (χ²_[79]  =  29.91; p > 0.99), so the CJS model could be used with our data. The specific tests for the CJS model assumption of equal catchability provided no evidence of a “trap dependence” effect for males (test 2.Ct, χ²_[13]  =  13.64, p  =  0.40) or females (test 2.Ct, χ²_[9]  =  2.46, p  =  0.98). Likewise, the specific tests for the assumption of homogeneous survival provided no evidence for the presence of transients among males (test 3.Sr, χ²_[10]  =  3.40, p  =  0.97) or females (test 3.Sr, χ²_[3]  =  0, p > 0.99). The assumption of homogeneous survival was not violated either for males (test 3.Sm, χ²_[6]  =  1.64, p  =  0.95) or females (test 3.Sm, χ²_[4]  =  0.63, p  =  0.96). Because the overdispersal value ĉ, calculated as the χ² of the overall model divided by its degree of freedom, was 0.38, we set the ĉ value as 1.00, not adjusting the AIC_c, and ran the models (White and Burnham 1999).

CJS model analysis indicated that the 3 best models had a similar power in explaining the encounter history data, because ΔAIC_c < 3 for each model (Burnham and Anderson 2002). The 3 models accounted for almost 100% of the Akaike weight (Table 2). Given the among-year variation in Cerrado environmental factors (Fig. 2A), models with capture probability that depended on the interaction of prior monthly rainfall and temperature, incorporated as constraints, represented 85% of the Akaike weight (Table 2, models 1 and 2). In addition, constraining the time variation of recapture probability to a monthly basis, i.e., when considering equal recapture probabilities for each month of the year, also contributed to model fit (Table 2, models 2 and 3). This indicates a recurrent pattern of recapture probabilities associated with the same month in different years.

View Figure

• Download as Powerpoint
• Download Figure

Table 2 Five best candidate models of Cormack-Jolly-Seber (CJS) model for monthly apparent survival (Φ) and monthly recapture probabilities (p) in Acanthochelys spixii. AIC_c  =  Akaike's information criterion for small samples (lower values meaning better model fit); ΔAIC_c  =  difference between the AIC_c of the current and the best model (model 1); WAIC_c  =  weight of fit for each model; K  =  number of estimated parameters. Subscripts: s, sex variation (males, females); t, time variation (i.e., 48 months); s*t, variation in both; rain and temp, variation in rainfall and average temperature of the current month; rain-1 and temp-1, variation in rainfall and average temperature of the previous month; month, monthly variation (i.e., equal survival or recapture probabilities for each month of the year); and (.), the parameter is held constant.

View Fullscreen

To incorporate model uncertainty in the parameter estimation, we calculated Φ and p under a model-averaging approach, which consisted of averaging parameters of candidate models weighted by their normalized AIC_c. Model averaging for monthly apparent survival (Φ) was 0.984, a constant value for both males and females (95% CI, 0.969–0.992; SE  =  0.005). Annual survival probability, calculated as the monthly survival estimate raised to the 12th power, was 0.823. Recapture probability (p) values were low and varied from 0.005 to 0.232, with similar patterns over the years. From May to September of every year, a time period that coincides with the dry season, we rarely captured a turtle, and p varied from 0.005 to 0.044 (Fig. 2B).

Population-size estimates under the JS model in module POPAN of program MARK, which assumes only 1 cohort, resulted in the best data fit by the model with constant survival and time-dependent recapture probability: Φ., p_t, pent_t. Because constraints in the probability of entrance of new individuals (pent) via both immigration and recruitment (birth) make no biological sense, this parameter of the JS model was assumed to vary freely over time. Average monthly population size was 30.4, with values that varied from 10 to 35 adult individuals (Fig. 2C).

Annual population growth rates estimated under the model-averaged AIC_c-ranked Pradel models resulted in 2 values: 1.37 ± 0.19 SE (95% CI, 1.01–1.74) for the 2005–2006 interval and 0.59 ± 0.08 SE (95% CI, 0.43–0.73) for the 2006–2007 interval. The top-ranked model incorporated survival varying with time and sex, and both recapture probability and population growth rate with time variation Φ_s*t, p_t, λ_t with 70% of the Akaike weight (Table 3).

Table 3 Five best candidate models for population growth rate estimation (λ) based on Pradel model analysis for Acanthochelys spixii. (See Table 2 for definition of abbreviations.)

View Fullscreen

DISCUSSION

Adult sex ratio represents a fundamental demographic parameter because the proportion of sexes may affect population dynamics (Lovich 1996). Unbalanced ratios are known for many wild turtle populations, as well as biased sampling associated with season of capture and trapping techniques (Gibbons 1990). Although temperature-dependent sex determination may also influence turtle sex ratios (Schwanz et al. 2010), all chelids probably have genotypic sex determination (Janzen and Paukstis 1991). Moreover, interpreting causes of adult turtle sex ratios depends on sufficient population records, but demographic data exist for few turtle species (Gibbons 1990). In natural populations of Emydura macquarii, adult sex ratios were strongly male biased in 9 of 11 populations examined and even an analysis adjustment with matched cohorts was insufficient to explain the strong male bias (1.2–2.9) in 5 of the 9 populations (Georges et al. 2006). In the longest demographic research conducted in Brazil for a chelid, a 1∶2 female-biased ratio was reported for Hydromedusa maximiliani in the Atlantic Forest, apparently driven by differential mortality and/or unknown natural factors (Martins and Souza 2009). As in our study, the sex ratio of Mesoclemmys hogei did not differ significantly from 1∶1 (Moreira 2002). In reptile populations with genotypic sex determination, a 1∶1 sex ratio is expected to occur (Janzen and Paukstis 1991). We believe that no sampling bias associated with season or trap methods affected our results, because our study was conducted continuously across seasons, and GOF testing for trap dependence was not significant for either sex. Because sex ratios may differ among years (Gibbons 1990), it is important to continue efforts to monitor changes in this population, especially when considering its small size.

CJS constrained models 1 and 2, which accounted for 85% of the Akaike weight, that assume that recapture probabilities depends on the interaction of previous monthly total rainfall and average air temperature. Moreover, among the 3 best models, 2 considered recapture probability estimates that vary on a monthly basis. This recurrent pattern of monthly recapture probability, as well as the power of constrained models based on climatic data, seems related to the very regular weather patterns of the biome. Indeed, the Cerrado has the most predictable climate among Brazilian biomes (Nimer 1989). Turtle activity patterns are often associated with climate, with rainfall and air temperature being the most influential parameters in many species (Souza 2004). The higher capture rates during the rainy season were probably associated with rising water levels. Rainfall in the previous month and water levels were significantly correlated and, as the pond continually rises during raining periods, odonate abundance and richness proportionally increase (Habib J. Fraxe Neto, pers. obs.). Odonates are the most important prey for A. spixii at the study site (Brasil et al. 2011 [this issue of CCB]).

Low values of capture probability, as in our estimates, could result from emigration, hibernation, idiosyncrasies of Cerrado turtle populations, or an interaction of these factors. The study site is limited to a lentic portion of about 4 ha. However, population spatial boundaries are generally vague and hard to define (Williams et al. 2002), and it is possible that individuals use habitats beyond the pond or migrate to other aquatic habitats nearby (see Souza 2004; Horta 2008). Observations that corroborate this hypothesis include 1) the population inhabits a temporary pond that was completely dry just before the 2003 rainy season; 2) after measurements, turtles released at the pond margins often sought refuge in terrestrial vegetation instead of going directly to the water, where they were initially captured; 3) occasionally, we found animals at least 500 m away from the pond. Another hypothesis is hibernation during colder months (Cerrado dry season), as recorded in other studies (Molina and Rocha 1990). Radiotelemetry and/or studies of gene flow may be useful to test these hypotheses (Harless et al. 2009). Another explanation for low recapture probabilities is idiosyncrasies of chelonian populations in the Cerrado. Relatively high numbers and ease of capture characterize turtle populations in other Brazilian biomes, such as Atlantic rainforest (Martins and Souza 2009) and Amazonia (Fachín-Terán et al. 2003). However, populations with sufficient numbers to provide robust demographic estimates are often hard to locate in the Cerrado (F.L. Souza, pers. comm.). Moreover, most A. spixii in the Cerrado are small or medium sized, with no economic importance for humans, and, because of that, it is even harder to obtain information about populations.

The 2 best CJS models assumed constant apparent survival (Φ), and model averaging for the annual parameter estimate was 0.823, with no between-sex variation during the study period. Despite the strong wet–dry seasonality of the Cerrado, there was no evidence for the influence of climate on survival, as recorded elsewhere (Converse et al. 2005). High levels of adult survival are particularly desirable in management plans of species with life histories that make them prone to overexploitation because of increased adult mortality and its resulting low probability of reproductive success (Congdon et al. 1993; Heppell 1998). In Terrapene ornata, high adult survival is necessary to attain 30+ years of longevity (Converse et al. 2005). However, some researchers argue that differences in exploitation vulnerability should be evaluated on a species by species basis, so that species with “fast” life histories may provide some level of adult harvest (Fordham et al. 2009) and that turtles may respond to changes in demographic processes in ways not predicted by common demographic models (Spencer and Janzen 2010). Chelodina rugosa, a fast-growing, early-maturing, and highly fecund species, subject to indigenous harvesting and pig predation, may compensate for low adult survival rates with elevated hatchling survival, smaller female size at maturity, and increased female postmaturity growth (Fordham et al. 2009).

The 2 values for annual population growth (λ), 1.37 for 2005–2006 and 0.59 for 2006–2007, must be interpreted with caution because of the short period of estimable time intervals. However, when considering the small population size, these estimates suggest that the population is either persisting at low abundance or declining in numbers. Therefore, continued estimation of population growth and size is warranted. The monthly population size curve (Fig. 2C) stabilized after August 2005 and remained at approximately 30 individuals. This figure is relatively small compared with 2 studies with chelids endemic to the Brazilian Atlantic rainforest. In São Paulo, the estimated abundance of Hydromedusa maximiliani was 235 (95% CI, 92–1016) from 1993–1994 and 318 (95% CI, 101–1842) from 2003–2006 (Martins and Souza 2009). In Minas Gerais, the population size of Mesoclemmys hogei at the Carangola River varied between 6 and 132 turtles from April 2001 to March 2002 (Moreira 2002). In contrast, Amazonian turtles are much more abundant: Fachín-Terán et al. (2003) reported 2458 captures of adult Podocnemis sextuberculata from September 1996 to August 1998 at Reserva de Desenvolvimento Sustentável Mamirauá. These numbers highlight the importance of finding and studying populations of chelids in Cerrado, given their relatively low abundance.

Predicting demographic estimates must account for habitats, because size and environmental and ecological aspects of habitat patches may influence demographic values (Huston 2002). An urban matrix isolates PNB, and anthropogenic pressures on this protected area include a large waste dump (Lixão da Estrutural), which attracts feral dogs, rats, and birds of prey that probably are responsible for depredation of at least 7 turtles at this study site since 2004. In C. rugosa and E. macquarii, which have predation by feral pigs and foxes, long-term population declines were not compensated even by higher adult survival rates (Fordham et al. 2006). Knowledge about habitat and its biota should be used to manage natural populations in a sustainable way (Huston 2002). As myriad fragments now distributed along vast portions of the original Cerrado landscape, the result of accelerated conversion into agricultural and urban areas, PNB is an island of biodiversity, and its reptile fauna may well resemble the original communities that lived therein (Rodrigues 2005). Our results indicate that the population of A. spixii inside PNB is relatively stable, within the limits of our sampling, therefore, this conservation unit is fulfilling its goals. The preservation of natural ecosystems inside PNB is pivotal for the maintenance of A. spixii in the Distrito Federal region.